Past Sessions

• Event: MAU Vegas 25
• Date: Thursday, May 21, 2025
• Speakers: Paul Gildard, Growth Creative, Copper, Michael Stenbakken, Chief Growth Officer, Copper
• Estimated read time: 5 to 6 minutes

Quick Read Summary

Platform automation has turned performance marketing creative into the most controllable growth lever left.

When targeting and bidding are increasingly handled by algorithms, the edge shifts to the ad, the message, the proof, and the pace of iteration. A control is simply the current winning creative that receives most of the spend. The job is to beat that control with a steady pipeline of concepts and variations, not to defend it as a permanent asset.

Michael Stenbakken and Paul Gildard share a practical playbook from Copper and earlier direct response work, focused on creative testing, trust building for financial claims, and faster production through selective use of AI.

Control is a moving target, build a system to beat it

Stenbakken defines the control as the creative currently taking the majority of budget, and he frames iteration as a disciplined attempt to outperform it. That framing keeps creative out of taste debates and puts it into a product like loop, ship, measure, learn, repeat.

The implication is planning for turnover. Controls decay as users fatigue and inventory shifts. If your process assumes one winner will last, you will end up replacing it late, with rushed bets and low signal.

The more durable approach is portfolio thinking. Keep several concepts in motion so a single control is never your only plan, and your iteration cadence stays steady.

“The control is your winning creative that the majority of your budget is going to, and when you are doing your creative iteration, you want to beat that so you can improve your performance.” Michael Stenbakken, role not stated in transcript, Copper

Start broad, find signal, then invest in polish

A case study in the transcript comes from ShoeDazzle, described as a Kim Kardashian Shoe of the Month Club era program running Facebook desktop right rail ads, spending about $40,000 per day. The team tested product images at high volume and learned that simple product shots beat images of Kim and other concepts. From that breadth testing, a specific shoe style became the control, Stenbakken recalls a red and black lacy shoe.

The lesson is sequence. Learn first, then produce. Stenbakken is clear that you can refine a winning concept and make it prettier, but you do not begin there. You begin with ads that are easier to produce so you can explore more concepts, then narrow into winners.

For senior leaders, this is an operating model choice. If you require high polish before you have proof, you slow iteration and reduce test volume. Separate learning assets from scaling assets and earn the right to polish by proving a concept can beat control.

If you make financial claims, trust comes from proof and obtainability

At Copper, the team runs rewarded gaming and makes financial claims, and Gildard calls out the problem directly, financial claims can feel untrustworthy. His answer is a stack built on trust, proof, and obtainability, making the outcome feel real and reachable.

He borrows a direct to consumer technique, tactile evidence. Copper applied the same principle to an app by showing a physical hand navigating screens, playing games, and demonstrating the reward process. The hand turns an abstract promise into a concrete experience.

They also take cues from YouTube style reviews, showing steps and milestones with transparency. Gildard adds a useful twist, inject skepticism and overcome it inside the same creative. Lines like “this sounds too good to be true” or “this will not make you a millionaire” make the message feel more like a review than a pitch.

User testing reinforced the need to design for distracted viewing, children, loud TVs, people in cars. Copper responded with simple language, repetition, and recap, repeating key claims at least twice and reinforcing them in captions for sound off viewing.

If your category triggers doubt, beat control by making belief easier than disbelief, show proof as a sequence, keep language plain, and design for partial attention.

Speed comes from looser constraints and smarter tooling

Stenbakken draws a line between brand work and performance work. Brand matters, but forcing full brand guidelines into performance campaigns can slow the learning loop. Performance teams need clear minimum guardrails, do nothing that harms the brand, but they also need autonomy to iterate.

Copper also builds speed by staying close to market examples, pulling ideas from ad libraries, recording ads from their own feeds, and sharing them internally. The goal is not novelty, it is a steady input stream for creative testing.

Gildard extends the speed theme with pragmatic AI use. Creator reads were costly and slow, so Copper used ElevenLabs to build a consistent voice from creator samples, then added humanizing modifiers such as just and now, plus deliberate imperfections like ums and coughs.

When Copper pivoted from a banking platform to rewarded gaming, they faced a backlog problem, top performing ads had working variables they wanted to keep, but reshoots were not feasible. They used AI again to update language and swap elements while preserving the strongest controls.

AI is most valuable here as a compression tool. It removes bottlenecks so your ads can adapt as fast as your product, without restarting the learning curve.

Design the screen, then build a repeatable formula

Gildard argues that ads should visually hypnotize and harmonize, meaning the viewer should not have to work to understand the message. His suggestion is practical, watch a top ad and notice how often your eyes have to move around the screen. Then pick a focal point and bring the key elements to that one spot.

He calls this optimal eye line and the rule is simple, do not let elements compete in the same shot. Faces carry heavy visual gravity, so if the key proof is on a phone screen, avoid adding a face or popups that steal attention. The goal is little to no movement so the message lands quickly.

This connects to structure. Gildard outlines a standard formula, thumb stopper, problem setup, solution or brand intro, value proposition, call to action. Copper layered in additional variables such as proof, a featured game section, cash out instructions, and more tactile evidence, while keeping the ad under one minute by keeping the layout harmonized.

On measurement, the transcript is candid. They use spreadsheets, but they also rely on familiarity with what they ship. The team treats early variables as most important, especially the thumb stopper, and will A B test multiple thumb stops and track them. They also run deeper reviews a few times a year, checking retention once there is enough time for data to bake.

The practical operating model is two loops, fast iteration on early creative variables, and periodic audits that connect those wins to longer term user quality.

Conclusion

Beating control is not a one time breakthrough, it is a repeatable system for producing new winners. As platforms automate more of the media stack, the teams that win will be the ones that can ship, learn, and refresh performance marketing creative faster than competitors, while keeping proof clear and trust intact.

Speakers

Paul Gildard, Growth Creative, Copper.  Based in Los Angeles, Paul specializes in data driven, performance focused video and D2C creative. His editing and design work spans brands including Copper, GilletteLabs, Native, Fabletics, Born Primitive, APL, BondiBoost, and Laura Geller.

Michael Stenbakken, Chief Growth Officer, Copper.  Marketing executive focused on profitable growth across consumer businesses, spanning acquisition, retention, offline and online channels, and growth strategy. Experienced in brand building, channel development, and product development, with a track record of building and leading high performing, results driven teams.